arXiv subjects

Yue Wu

Publications and source records attributed to Yue Wu.

At least 397 records · Page 22Linked to original sources

arXiv2017

Constrained Joint Cascade Regression Framework for Simultaneous Facial Action Unit Recognition and Facial Landmark Detection

Cascade regression framework has been shown to be effective for facial landmark detection. It starts from an initial face shape and gradually predicts the face shape update from the local appearance features to generate the facial landmark locations in the next iteration until convergence. In this paper, we improve upon the cascade regression framework and propose the Constrained Joint Cascade Regression Framework (CJCRF) for simultaneous facial action unit recognition and facial landmark detection, which are two related face analysis tasks, but are seldomly exploited together. In particular, we first learn the relationships among facial action units and face shapes as a constraint. Then, in the proposed constrained joint cascade regression framework, with the help from the constraint, we iteratively update the facial landmark locations and the action unit activation probabilities until convergence. Experimental results demonstrate that the intertwined relationships of facial action units and face shapes boost the performances of both facial action unit recognition and facial landmark detection. The experimental results also demonstrate the effectiveness of the proposed method comparing to the state-of-the-art works.

cs.CV↗

arXiv2017

Simultaneous Facial Landmark Detection, Pose and Deformation Estimation under Facial Occlusion

Facial landmark detection, head pose estimation, and facial deformation analysis are typical facial behavior analysis tasks in computer vision. The existing methods usually perform each task independently and sequentially, ignoring their interactions. To tackle this problem, we propose a unified framework for simultaneous facial landmark detection, head pose estimation, and facial deformation analysis, and the proposed model is robust to facial occlusion. Following a cascade procedure augmented with model-based head pose estimation, we iteratively update the facial landmark locations, facial occlusion, head pose and facial de- formation until convergence. The experimental results on benchmark databases demonstrate the effectiveness of the proposed method for simultaneous facial landmark detection, head pose and facial deformation estimation, even if the images are under facial occlusion.

cs.CV↗

arXiv2017

Facial Feature Tracking under Varying Facial Expressions and Face Poses based on Restricted Boltzmann Machines

Facial feature tracking is an active area in computer vision due to its relevance to many applications. It is a nontrivial task, since faces may have varying facial expressions, poses or occlusions. In this paper, we address this problem by proposing a face shape prior model that is constructed based on the Restricted Boltzmann Machines (RBM) and their variants. Specifically, we first construct a model based on Deep Belief Networks to capture the face shape variations due to varying facial expressions for near-frontal view. To handle pose variations, the frontal face shape prior model is incorporated into a 3-way RBM model that could capture the relationship between frontal face shapes and non-frontal face shapes. Finally, we introduce methods to systematically combine the face shape prior models with image measurements of facial feature points. Experiments on benchmark databases show that with the proposed method, facial feature points can be tracked robustly and accurately even if faces have significant facial expressions and poses.

cs.CV↗

arXiv2017

A Hierarchical Probabilistic Model for Facial Feature Detection

Facial feature detection from facial images has attracted great attention in the field of computer vision. It is a nontrivial task since the appearance and shape of the face tend to change under different conditions. In this paper, we propose a hierarchical probabilistic model that could infer the true locations of facial features given the image measurements even if the face is with significant facial expression and pose. The hierarchical model implicitly captures the lower level shape variations of facial components using the mixture model. Furthermore, in the higher level, it also learns the joint relationship among facial components, the facial expression, and the pose information through automatic structure learning and parameter estimation of the probabilistic model. Experimental results on benchmark databases demonstrate the effectiveness of the proposed hierarchical probabilistic model.

cs.CV↗

arXiv2017

A randomized Milstein method for stochastic differential equations with non-differentiable drift coefficients

In this paper a drift-randomized Milstein method is introduced for the numerical solution of non-autonomous stochastic differential equations with non-differentiable drift coefficient functions. Compared to standard Milstein-type methods we obtain higher order convergence rates in the $L^p(Ω)$ and almost sure sense. An important ingredient in the error analysis are randomized quadrature rules for Hölder continuous stochastic processes. By this we avoid the use of standard arguments based on the Itō-Taylor expansion which are typically applied in error estimates of the classical Milstein method but require additional smoothness of the drift and diffusion coefficient functions. We also discuss the optimality of our convergence rates. Finally, the question of implementation is addressed in a numerical experiment.

math.NA↗

arXiv2017

Deep Matching and Validation Network -- An End-to-End Solution to Constrained Image Splicing Localization and Detection

Image splicing is a very common image manipulation technique that is sometimes used for malicious purposes. A splicing detec- tion and localization algorithm usually takes an input image and produces a binary decision indicating whether the input image has been manipulated, and also a segmentation mask that corre- sponds to the spliced region. Most existing splicing detection and localization pipelines suffer from two main shortcomings: 1) they use handcrafted features that are not robust against subsequent processing (e.g., compression), and 2) each stage of the pipeline is usually optimized independently. In this paper we extend the formulation of the underlying splicing problem to consider two input images, a query image and a potential donor image. Here the task is to estimate the probability that the donor image has been used to splice the query image, and obtain the splicing masks for both the query and donor images. We introduce a novel deep convolutional neural network architecture, called Deep Matching and Validation Network (DMVN), which simultaneously localizes and detects image splicing. The proposed approach does not depend on handcrafted features and uses raw input images to create deep learned representations. Furthermore, the DMVN is end-to-end op- timized to produce the probability estimates and the segmentation masks. Our extensive experiments demonstrate that this approach outperforms state-of-the-art splicing detection methods by a large margin in terms of both AUC score and speed.

cs.CV↗

arXiv2017

Families in the Wild (FIW): Large-Scale Kinship Image Database and Benchmarks

We present the largest kinship recognition dataset to date, Families in the Wild (FIW). Motivated by the lack of a single, unified dataset for kinship recognition, we aim to provide a dataset that captivates the interest of the research community. With only a small team, we were able to collect, organize, and label over 10,000 family photos of 1,000 families with our annotation tool designed to mark complex hierarchical relationships and local label information in a quick and efficient manner. We include several benchmarks for two image-based tasks, kinship verification and family recognition. For this, we incorporate several visual features and metric learning methods as baselines. Also, we demonstrate that a pre-trained Convolutional Neural Network (CNN) as an off-the-shelf feature extractor outperforms the other feature types. Then, results were further boosted by fine-tuning two deep CNNs on FIW data: (1) for kinship verification, a triplet loss function was learned on top of the network of pre-trained weights; (2) for family recognition, a family-specific softmax classifier was added to the network.

cs.CV↗

arXiv2017

Red Clump stars from LAMOST II: the outer disc of the Milky Way

We present stellar density maps of the Galactic outer disc with red clump stars from the LAMOST data. These samples are separated into younger (mean age ~ 2.7 Gyr) and older (mean age ~ 4.6 Gyr) populations so that they can trace the variation of the structures with ages in the range of the Galactocentric radius R from 9 to 13.5 kpc. We show that both the scale heights for the two populations increase with R and display radial gradients of 48 +/- 6 and 40 +/- 4 pc/kpc for the older and younger populations, respectively. This is evident that the flaring occurs in the thin disc populations with a wide range of ages. Moreover, the intensity of flaring seems not significantly related to the age of the thin disc populations. On the other hand, the scale lengths of the radial surface density profiles are 4.7 +/- 0.5 kpc for the younger and 3.4 +/- 0.2 kpc for the older population, meaning that the younger disc population is more radially extended than the older one. Although the fraction of the younger population mildly increases from 28% at R ~ 9 to about 35% at R ~ 13 kpc, the older population is prominent with the fraction no less than 65% in the outer disc.

astro-ph.GA↗

arXiv2017

Candidate members of the Pal 5, GD-1, Cetus Polar, and Orphan tidal stellar halo streams from Sdss DR9, Lamost DR3 and APOGEE catalogues

We present candidate members of the Pal 5, GD-1, Cetus Polar, and Orphan tidal stellar streams found in LAMOST DR3, SDSS DR9 and APOGEE catalogs. In LAMOST DR3, we find 20, 4, 24 high confidence candidates of tidal streams GD-1, Cetus Polar and Orphan respectively. We also list from the SDSS DR9 spectroscopic catalog 59, 118, 10 high confidence candidates of tidal streams Cetus Polar, Orphan and Pal 5, respectively. Furthermore, we find 7 high confidence candidates of the Pal 5 tidal stream in APOGEE data. Compared with SDSS, the new candidates from LAMOST DR3 are brighter, so that together, more of the color-magnitude diagram, including the giant branch can be explored. Analysis of SDSS data shows that there are 3 metallicity peaks of the Orphan stream and also shows some spatial separation. LAMOST data confirms multiple metallicities in this stream. The metallicity, given by the higher resolution APOGEE instrument, of the Pal 5 tidal stream is [Fe/H] $\sim -1.2$, higher than that given earlier by SDSS spectra. Many previously unidentified stream members are tabulated here for the first time, along with existing members, allowing future researchers to further constrain the orbits of these objects as they move within the Galaxy's dark matter potential.

astro-ph.GA↗

arXiv2017

Error analysis of randomized Runge-Kutta methods for differential equations with time-irregular coefficients

This paper contains an error analysis of two randomized explicit Runge-Kutta schemes for ordinary differential equations (ODEs) with time-irregular coefficient functions. In particular, the methods are applicable to ODEs of Carathéodory type, whose coefficient functions are only integrable with respect to the time variable but are not assumed to be continuous. A further field of application are ODEs with coefficient functions that contain weak singularities with respect to the time variable. The main result consists of precise bounds for the discretization error with respect to the $L^p(Ω;\mathbb{R}^d)$-norm. In addition, convergence rates are also derived in the almost sure sense. An important ingredient in the analysis are corresponding error bounds for the randomized Riemann sum quadrature rule. The theoretical results are illustrated through a few numerical experiments.

math.NA↗

arXiv2017

The local spiral arm in the LAMOST-Gaia common stars?

Using the LAMOST-Gaia common stars, we demonstrate that the in-plane velocity field for the nearby young stars are significantly different from that for the old ones. For the young stars, the probably perturbed velocities similar to the old population are mostly removed from the velocity maps in the $X$--$Y$ plane. The residual velocity field shows that the young stars consistently move along $Y$ with faster $v_ϕ$ at the trailing side of the local arm, while at the leading side, they move slower in azimuth direction. At both sides, the young stars averagely move inward with $v_R$ of $-5\sim-3$ km s$^{-1}$. The divergence of the velocity in $Y$ direction implies that the young stars are associated with a density wave nearby the local arm. We therefore suggest that the young stars may reflect the formation of the local spiral arm by correlating themselves with a density wave. The range of the age for the young stars is around 2 Gyr, which is sensible since the transient spiral arm can sustain that long. We also point out that alternative explanations of the peculiar velocity field for the young population cannot be ruled out only from this observed data.

astro-ph.GA↗

arXiv2016

SOL: A Library for Scalable Online Learning Algorithms

SOL is an open-source library for scalable online learning algorithms, and is particularly suitable for learning with high-dimensional data. The library provides a family of regular and sparse online learning algorithms for large-scale binary and multi-class classification tasks with high efficiency, scalability, portability, and extensibility. SOL was implemented in C++, and provided with a collection of easy-to-use command-line tools, python wrappers and library calls for users and developers, as well as comprehensive documents for both beginners and advanced users. SOL is not only a practical machine learning toolbox, but also a comprehensive experimental platform for online learning research. Experiments demonstrate that SOL is highly efficient and scalable for large-scale machine learning with high-dimensional data.

cs.LG↗

arXiv2016

Nano-Cross-Junction Effect on Phonon Transport in Silicon-Nanowire-Cages

Wave effects of phonons can give rise to controllability of heat conduction beyond that by particle scattering at surfaces and interfaces. In this work, we propose a new class of 3D nanostructure: a silicon-nanowire-cage (SiNWC) structure consisting of silicon nanowires (SiNWs) connected by nano-cross-junctions (NCJs). We perform equilibrium molecular dynamics (MD) simulations, and find an ultralow value of thermal conductivity of SiNWC, 0.173 Wm-1K-1, which is one order lower than that of SiNWs. By further modal analysis and atomistic Green's function calculations, we identify that the large reduction is due to significant phonon localization induced by the phonon local resonance and hybridization at the junction part in a wide range of phonon modes. This localization effect does not require the cage to be periodic, unlike the phononic crystals, and can be realized in structures that are easier to synthesize, for instance in a form of randomly oriented SiNWs network.

cond-mat.mes-hall↗

arXiv2016

Exoplanet Orbital Eccentricities Derived From LAMOST-Kepler Analysis

The nearly circular (mean eccentricity ~0.06) and coplanar (mean mutual inclination ~3 deg) orbits of the Solar System planets motivated Kant and Laplace to put forth the hypothesis that planets are formed in disks, which has developed into the widely accepted theory of planet formation. Surprisingly, the first several hundred extrasolar planets (mostly Jovian) discovered using the Radial Velocity (RV) technique are commonly on eccentric orbits ( ~ 0.3). This raises a fundamental question: Are the Solar System and its formation special? The Kepler mission has found thousands of transiting planets dominated by sub-Neptunes, but most of their orbital eccentricities remain unknown. By using the precise spectroscopic host star parameters from the LAMOST observations, we measure the eccentricity distributions for a large (698) and homogeneous Kepler planet sample with transit duration statistics. Nearly half of the planets are in systems with single transiting planets (singles), while the other half are multiple-transiting planets (multiples). We find an eccentricity dichotomy: on average, Kepler singles are on eccentric orbits with ~0.3, while the multiples are on nearly circular ( = 0.04^{+0.03}_{-0.04}) and coplanar ( = 1.4^{+0.8}_{-1.1} deg) orbits similar to the Solar System planets. Our results are consistent with previous studies of smaller samples and individual systems. We also show that Kepler multiples and solar system objects follow a common relation ~(1-2)x between mean eccentricities and mutual inclinations. The prevalence of circular orbits and the common relation may imply that the solar system is not so atypical in the galaxy after all.

astro-ph.EP↗

arXiv2016

The qualitative analysis of the impact of media delay on the control of infectious disease

In this paper, we consider the impact of time delay by media on the control of the disease. We set up a class of SISM epidemic model with the time delay and the cumulative density of awareness caused by media. The sufficient condition of global asymptotic stability of disease-free equilibrium is approved. We get the global stability of the epidemic equilibrium and the existence conditions of Hopf bifurcation. Numerical simulations are presented to illustrate the analytical results. Finally, we analyze the influence of parameters on the control of infectious disease by combining the data of H1N1. By shortening the time of media lag, increasing transmission rate of media and the implementation rate of the media project, the spread of disease will be controlled effectively.

q-bio.PE↗

arXiv2016

The Dynamic Behavior (Effects) of Impulsive Toxicant Input on a Single-Species Population in a Small Polluted Environment

In this paper, we study a single-species population model with pulse toxicant input to a small polluted environment. The intrinsic rate of population is affected by environment and toxin in organisms. The toxin in organisms is influenced by toxin in environment and the food chain. A new mathematical model is derived. By the Pulse Compare Theorem, we find the surviving threshold of the population and obtain the sufficient conditions of persistence and extinction of the population.

q-bio.PE↗

arXiv2016

A Family of Totally Rank One Two-sided Shift Maps

'Generalized del Junco-Rudolph's map', a sub-family of generalized Chacon's map (\cite{FER1}), is introduced. A skew product related to the structure of the Generalized del Junco-Rudolph's map is introduced. A Relative Prime Relation, $h_{k+1}\equiv 1 \mod q$ is verified, based on the proposition of iterations of this skew product. We say a measure-preserving transformation $T$ is totally rank one if $T^{q}$ is rank one for every $q\in \mathbb{N}$. In this paper, we show that of every Generalized del Junco-Rudolph's map is totally rank one.

math.DS↗

arXiv2016

Totally Rank One Interval Exchange Transformations

For irreducible interval exchange transformations, we study the relation between the powers of induced map and the induced maps of powers and raise a condition of equivalence between them. And skew production of Rauzy induction map is set up and verified to be ergodic regard to a product measure. Then we prove that almost all the interval exchange transformations are totally rank one (rank one for all powers of positive integers) by interval. As a corollary, for almost all interval exchange transformations, rank one transformations are dense $G_δ$ in the weak closure.

math.DS↗